Time to first recurrence as a trial endpoint: time to change?

There has been a rapidly increasing interest in developing new therapies for management of atrial fibrillation. The optimal method for evaluating their efficacy is by measuring changes in outcomes such as mortality, quality of life or cost of care; requiring trials of large sample size. In order to reduce the sample size, there is a need to develop appropriate surrogate endpoints. "Is time to first recurrence of symptomatic atrial fibrillation" an appropriate surrogate endpoint for quality of life? Since a surrogate endpoint must capture the net effect of the treatment, it was assumed (a) that frequency of symptomatic episodes captures the net effect on quality of life and (b) "time to first recurrence" is a measure of the frequency of symptomatic episodes. The effect of frequency of symptomatic episodes or their duration and symptom severity on quality of life has not been evaluated. "Time to first symptomatic recurrence" was proposed because data from a few patients demonstrated that symptomatic atrial fibrillation episodes arose independently and randomly and could be represented mathematically by a Poisson distribution. Recent data from a greater number of patients with implantable devices that detect symptomatic and asymptomatic episodes of atrial tachyarrhythmias indicate that these episodes tend to cluster in time and cannot be well represented by a Poisson distribution. Because all atrial tachyarrhythmia episodes do not follow a Poisson distribution, it is unlikely that only symptomatic episodes would follow the same distribution, although this needs to be proved. A non-Poisson distribution requires a larger sample to detect differences in clinical trials. This should be taken into consideration when designing prospective trials. Alternative methods may include measuring frequency of symptomatic episodes as well their severity and duration. In patients with implanted devices, total duration of time spent in atrial tachyarrhythmias or objective measures of rate control should also be evaluated as surrogate endpoints. If surrogate endpoints cannot be developed, changes in clinical outcomes will need to be demonstrated to evaluate therapeutic efficacy.